JST Vol. 21 (1) Jan. 2013 - Pertanika Journal - Universiti Putra ...

Results and Analysis
Yuhanis Yusof and Mohammed Hayel Refai
Table 3 contains the results of the classification accuracy, while data in Table 4 depicts the
number of association rules produced by MCAR and MMCAR. Data in Table 4 show similar
consistency in the classification accuracies of both MMCAR and MCAR. The accuracy values
obtained by both MCAR and MMCAR for six data sets (Austra, Balance-scale, Labor, Lymph,
Mushroom and Wine) are at the same level; they only differ at the decimal point. For example,
using Austra dataset, both classifiers generate 86% accuracy. On the other hand, MMCAR
outperformed the accuracy obtained by MCAR in five data sets. Such a scenario is illustrated
in Fig.4, where it shows the difference of accuracy values between MMCAR and MCAR. For
example, the difference is as high as 2.85% for the Glass dataset, while there is no difference
for the Labour dataset.
TABLE 3: Prediction Accuracy
Data set MCAR MMCAR
Austra 86.14 86.26
Balance-scale 76.96 76.17
Breast 94.99 93.83
Cleve 81.84 78.77
Glass 71.35 74.2
Heart-s 81.15 80.51
Iris 92.93 94.26
Labor 83.5 83.5
Led7 71.83 73
Lymph 78.1 78.05
Mushroom 99.6 99.67
Pima 77.12 74.44
Vote 88.2 86.39
Wine 95.73 95.73
Fig.4: The Difference in Prediction Accuracy
212 Pertanika J. Sci. & Technol. 21 (1): 283 - 298 (2013)